The other time-curves, C and D, corresponding to the initial and maximum epochs of the third phase, are practically straight lines. Some of the records are slightly discordant for the average curve, especially for the initial epoch; but it is often difficult to define the commencement of this phase with precision. At any rate, the observations show no distinct sign of an increase in the surface-velocity of these waves with the distance from the origin. It may therefore be concluded that they travel along the surface with velocities which are practically constant for each individual earthquake, the largest waves at the rate of about 2.9 kms. per sec., and the advance waves with a velocity of about 3.3 kms. per sec., rising occasionally to over 4.0 kms. per sec. STRUCTURAL CHANGES IN THE EPICENTRAL AREA. Changes of elevation have long been known as accompaniments of great earthquakes, though many of the earlier observations and measurements left much to be desired in accuracy and completeness. The Japanese earthquake of 1891, however, placed the reality of such movements beyond doubt, and revealed the existence of a fault-scarp, with a height in one place of 18 or 20 feet, and a length of at least 40, if not of 70, miles. In the Indian earthquake of 1897, the fault-scarps were shorter, though more pronounced in character, the largest known (the Chedrang fault) being about 12 miles long, and having a maximum throw at the surface of 35 feet. In some other recent earthquakes, also, remarkable fault-scarps have been developed. After the great shocks felt in Eastern Greece on April 20th and 27th, 1894, a fissure was traced for a distance of about 34 miles, running in an east-south-east and west-north-west direction through the epicentral district, and varying in width from an inch or two to more than three yards. That it was a fault, and not an ordinary fissure, was evident from its great length, its uniform direction, and its independence of geological structure. The throw was generally small, in no place exceeding five feet.[89] Again, in British Baluchistan, after the severe earthquake of December 20th, 1892, a fresh crack was observed in the ground running for several miles in a straight line parallel to the axis of the Khojak range. It coincided almost exactly with a line of springs, and was clearly produced by a fresh slip along an old line of fault, for before the earthquake it had the appearance of an old road, and the natives assert that the ground has always cracked along this line with every severe shock. In 1892, the change in relative height of the two sides of the fault was small, in one place where it was measured being only two inches.[90]